Toward Principled Transformers for Knowledge Tracing
Keywords: educational data mining, knowledge tracing, transformer
TL;DR: We propose knowledge tracing set transformers, a straightforward model class for knowledge tracing that is conceptually simpler than previous state-of-the-art approaches while outperforming them on standardized benchmark datasets.
Abstract: Knowledge tracing aims to reason about changes in students' knowledge and to predict students' performance in educational learning settings. We propose knowledge tracing set transformers (KTSTs), a straightforward model class for knowledge tracing prediction tasks. This model class is conceptually simpler than previous state-of-the-art approaches, which are overly complex due to domain-inspired components, and which are in part based on suboptimal design choices and flawed evaluation. In contrast, for KTSTs we propose principled set representations of student interactions and a simplified variant of learnable modification of attention matrices for positional information in a student's learning history. While being largely domain-agnostic, the proposed model class thus accounts for characteristic traits of knowledge tracing tasks. In extensive empirical experiments on standardized benchmark datasets, KTSTs establish new state-of-the-art performance.
Supplementary Material: zip
Primary Area: other topics in machine learning (i.e., none of the above)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 6646
Loading